THE EFFECTS OF TILLAGE, GLYPHOSATE, AND GENETIC MODIFICATION ON

BACTERIAL ROOT ENDOPHYTE COMPOSITION IN ZEA MAYS

by

Breena Lori Frieda Nolan

A thesis submitted to the faculty of the University of Mississippi in partial fulfillment of the

requirements of the Sally McDonnell Barksdale Honors College.

Oxford

November 2016

Approved by

Advisor: Dr. Colin Jackson

Reader: Dr. Michael Jenkins

Reader: Dr. John Samonds

© 2016

Breena Lori Frieda Nolan

ALL RIGHTS RESERVED

ii

ACKNOWLEDGEMENTS

Many thanks to USDA-ARS for supplying the experimental design, technical knowledge,

and materials for this experiment, as well as preparing the root material for processing,

particularly Dr. Michael Jenkins and Dan McChesney. Thanks as well for the financial support

for this experiment provided by the Sally Barksdale Honors College and USDA-ARS. Finally,

many thanks to Dr. Colin Jackson, who refined the experimental design, provided all training

with mothur and the associated programs, and advised throughout the production of this thesis.

It is thanks to their efforts that this paper was written.

iii

ABSTRACT

The community structure of the endophytic bacteria in Zea mays roots was examined for

the potential effects of glyphosate application, tillage strategies, and whether or not the corn

plant in question was of an organic or glyphosate-resistant variety. Roots were harvested from

plots designated to receive their specific treatments at the USDA-ARS Crop Production Systems

Research Unit Farm. Vortexing, sonication, and tissue grinding, extraction, and next generation

sequencing of 16S rRNA genes from these roots were used to describe their bacterial community

composition. Results indicated significant differences in the bacterial communities correlated to

tillage practice or corn type, whereas glyphosate treatments did not seem to affect the bacterial

community. There also appeared to be certain holistic differences resulting from the

combinations of certain treatments. Prior research has focused primarily on fungal endophytes,

but as 16S rRNA sequencing has immeasurably broadened the scope of microbiological studies,

new research such as this seeks to identify new microbes and their potential functions in the

macroscopic world.

iv

TABLE OF CONTENTS

LIST OF TABLES……………………………………………………………………......... vi

INTRODUCTION………………………………………………………………………….. 1

METHODS…………………………………………………………………………………. 5

RESULTS AND DISCUSSION……………………………………………………………. 11

CONCLUSION…………………………………………………………………………….. 20

BIBLIOGRAPHY………………………………………………………………………….. 22

v

LIST OF TABLES

Table 1 Commands used in the bioinformatics software mothur to process the

16S rRNA gene sequence data obtained in this study…………………. 10

Table 2 The fifteen most significant OTUs, as calculated by NMDS, OTU

abundance, and the Spearman’s rank correlation coefficient………….. 15

Figure 1 NMDS plot of the bacteria that appeared most responsible for the

greatest differentiation between samples……………………………… 15

Table 3 The p-value results as obtained through mothur’s AMOVA analysis

while comparing the following variables……………………………… 17

Table 4 Indicator OTUs displayed with their p-values, identities as far as the

greengenes database could classify their sequences, and the samples

they were significantly greater within…………………………………. 19

vi

INTRODUCTION

Over the last few decades, advances in DNA sequencing technology have led to

previously unexplored avenues of research in microbiology. Many microorganisms were

impossible to identify prior to 16S rRNA and rDNA analysis methods [13], resulting in extensive

gaps in our knowledge of the diversity studies of virtually all microbial communities. Between

2001 and 2007, the use of 16S rDNA sequencing identified 215 new bacterial species from

human specimens alone [18]. Although the library of known rRNA sequences is far from

complete, 106 rRNA sequences have been identified as of 2010, providing for 109 distinct 16S

rRNA gene sequence tags [13], which allow the classification and study of novel microbes that

have never been cultured, are rare, or are particularly slow-growing. Once an unknown

microorganism has been isolated, today’s molecular phylogeny techniques can be used to

determine its relatives in an effort to better understand these new specimens’ function in the

microbial community [14]. Sequencing techniques themselves have also been improving

rapidly. The development of pyrosequencing has allowed for the identification of nearly 100-

fold more sequences than the traditional Sanger method [12]. The Illumina HiSeq and MiSeq

platforms offer affordable, high-throughput sequencing, greatly expanding the range of feasible

microbe studies and experiments [2].

Among these newer areas of study are those that focus on niches that were previously

difficult to access without disrupting or killing the bacteria in question; for instance, the internal

tissues of plants. These limitations are in part responsible for the classic definition of an

endophyte as “fungi colonizing living plant tissue without causing any immediate, overt negative

effects” [6]. This overlooks the ubiquitous prevalence of bacterial endophytes in every plant,

7

and furthermore disregards that these symbiotic relationships may be a major component of that

plant’s growth and survival [17]. Endophytes encompass the full spectrum of symbiotic

interactions, with the only true defining feature of this group being the ability to live inside the

plant host’s tissue without killing its host [17]. As a reservoir of genetic diversity that likely

contains many undiscovered species, bacterial endophytes may offer extensive insight into

microbial diversity and phylogeny [17]. Model research systems of endophytes may also lead to

a broader understanding of plant-pathogen interactions and evolution.

Interest in prokaryotic endophytes has largely developed in agriculture, where

interactions between endophytes, the plant, and the broader environment have economic impacts.

For example, the nitrogen-fixing bacterium Acetobacter diazotrophicus, an endophyte of

sugarcane, allows the crop to be grown for long periods of time without the need to replenish soil

nitrogen through fertilizers [1]. This suggests that an improved understanding of how human

cultivation practices impact the microbial community and vice versa may in turn lead to higher

crop yields, hardier plants, and reduced capital lost on fertilization and pesticides. Likewise,

parasitic and disease-causing microorganisms might be easier to combat once a plant species’

internal microbial community is better understood.

A cultivation practice that is currently being examined is soil tillage, which can affect the

soil microbial community and lead to changes in nitrogen, phosphorus, and carbon cycling. In

turn, these can result in variations in plant growth, as well as affect the competitive inhibition of

plant pathogens. Reduced tillage, as opposed to conventional tillage, typically increases the

amount of carbon in the soil and total bacterial biomass in upper soil horizons, without affecting

bacterial growth rate [3, 5]. However, prior tillage studies have never taken into account the

8

specific bacterial community composition, nor have they considered the endophytes in the crops

grown in such systems.

As well as tillage practices, effects of pesticides on soil and endophyte bacterial

communities are also of interest, particularly the effects of glyphosate, N-

(phosphonomethyl)glycine, the active ingredient in Roundup herbicide [9]. Roundup is a post-

emergence, non-selective herbicide that works by inhibiting the enzyme 5-enolpyruvyl-

shikimate-3-phosphate synthase in the shikimate pathway [4], which would normally create

aromatic amino acids essential to protein and secondary metabolite synthesis in plants [9],

leading to a quick death upon exposure. A high uptake rate in plants, little degradation by plant

metabolism, low mobility in soil and groundwater, as well as being reportedly relatively non-

toxic to animals and non-carcinogenic [4], are traits that have made the use of Roundup

widespread. However, glyphosate causes enzymatic and reproductive disruptions in animals,

including a study that showed human placental cells begin to sustain damage at Roundup

concentrations 10 times lower than those used in agriculture, with the effect increasing over time

[11]. Roundup also directly inhibits aromatase activity in human microsomes, though pure

glyphosate had a much lessened effect as opposed to Roundup itself [11]. Glyphosate’s major

degradation product, aminophosphonic acid is much more mobile in the soil and might also have

an impact on microbial communities [4].

Another cultivation practice that relates to herbicide use is the increased production of

genetically modified plants. One example, glyphosate resistance, is seen in transgenic

glyphosate-resistant (GR) soybeans, which, when healthy, are able to metabolize glyphosate by

means of glyphosate oxidoreductase (GOX). Resistance can also be developed by insertion of

the CP4 gene of Agrobacterium into the desired plant genome [4], causing the production of GR

9

5-enolpyruvyl-shikimate-3-phosphate synthase. Knowing how the bacterial endophytes of GR

plants compare to those within their unmodified originals could provide information on the

broader effects of genetically modified crops.

Tillage, glyphosate application, and GR plants are interrelated practices. Glyphosate

treatments allow reduced tillage more easily by controlling weeds, and this tillage practice is

considered desirable due to its preservation of the top soil, prevention of the pollution of surface

waters and air, and the indirect reduction of carbon dioxide emissions [4]. However, to use

glyphosate, GR crops are required. Knowing how each of these cultivation practices affects the

endophytic community both individually and in conjunction may prove to be of incalculable

benefit to the agricultural community.

The aim of this research was to provide an initial analysis of the endophytes found in Zea

mays roots, with a particular emphasis on whether endophyte composition would be affected by

tillage methods (conventional versus reduced), pesticide exposure (glyphosate treatment versus

control), and genetic modification (GR plants versus control plants). This research was part of a

broader ARS-USDA study examining these treatments in a larger environmental context.

Genetic material was extracted from Zea mays root samples and portions of bacterial 16S rRNA

were examined using next generation sequencing. Results suggested only minimal influence of

glyphosate upon the microbial community, while tillage method and genetic modification

appeared to substantially alter the bacterial endophyte community composition.

10

METHODS

Study Site and Treatments

All samples were collected in 2014 from the USDA-ARS Crop Production Systems

Research Unit Farm near Stoneville, MS. Plots consisted of eight 32m long rows spaced 1m

apart. Three variables were tested with four replications: tillage (conventional versus reduced),

glyphosate (treatment versus no exposure), and Zea mays genetic modification (transgenic GR

plants versus non-GR). Field preparation was carried out via disking, subsoiling, disking, and

bedding in 2014. Conventionally tilled plots were subsoiled and prepared following a corn

harvest each fall, while reduced tillage plots were not tilled after the fall of 2014. Weeds were

eliminated either by herbicide application or hand hoeing. Herbicide application followed a

strict routine for every plot throughout the duration of the experiment: in February, all plots were

burned down with 2, 4-D (1.1kg ha-1), before planting, paraquat (2.2kg ha-1) was applied, and

immediately after planting, atrazine (1.7kg ha-1) and metolachor (1.7kg ha-1) were applied.

Glyphosate-treated plots received 2.2kg ha-1 applied twice during the early and late crop seasons

respectively. The early crop season application was sprayed over the top and the late crop

season application was applied directly to the base of the plant. No-exposure plots received no

glyphosate at any point. In plots with non-GR corn, glyphosate application was administered via

hooded sprayer between corn rows to avoid killing those plants. To manage yellow nutsedge,

halosufuron (0.07kg ha-1) was applied in the third week of May. All plots additionally received a

mixture of liquid urea and ammonium nitrate, providing them with 225kg N.

11

Root Collection and Processing

Using a flame-sterilized shovel, seven root balls were removed from each plot, beaten

against the shovel to shake off excess dirt, and stored in a paper bag for transit (time

unspecified). Once transferred to the lab, root balls from the same plot were then placed in a

bucket with 4L of sterile MilliQ water one after another and agitated gently to remove soil. The

root balls were washed with distilled water in a separate container to reduce soil further, at which

point a portion of root material was removed and stored in Ziploc bags at -80⁰C. Bags were

labelled with designations to clarify the tillage type, pesticide treatment, and corn type each root

sample represented.

Root Preparation

Frozen root samples were thawed, weighed, and 2.0g of 1-2cm cuts of root placed into a

50mL tared Falcon tube containing 25mL of Silwet buffer (composition being 5.7g NaH2PO4 x

H2O, 12.38g Na2HPO4 x 7H2O, and 150mL Silwet L-77 in 750mL MilliQ water). These tubes

were vortexed to remove soil from roots, which were transferred to fresh Falcon tubes containing

25mL of Silwet buffer. The vortexing and transference steps were repeated an average of five

times until all visible soil was dislodged from roots. Tubes were then sonicated at 40W output

for 30 seconds in a Cole Parmer 4710 Series Model CP 100 Ultrasonic Homogenizer, then

placed on ice for 1 minute to dislodge microbes from the root surface. The sonication and ice

steps were repeated five times each. Roots were again transferred to a fresh, empty Falcon tube,

while being clipped into smaller 0.5 cm lengths. The roots were left undisturbed in the Falcon

12

tube for 15 minutes to allow excess Silwet buffer to collect in the tube. Roots were then

removed and stored in a clean, dry Falcon tube at -80⁰C.

At all points during this procedure, roots were handled with flame-sterilized forceps, and

all clipping was carried out by flame-sterilized scissors. The sonication probe was wiped after

each root sample’s sonication step was completed to remove contaminants, after which followed

90% ethanol sterilization.

Root Disruption

Root samples were removed from -80⁰C and using flame-sterilized tweezers,

approximately half of the root sample was immediately transferred into a clean, sterile tissue

grinder. The other half of root sample was stored at -20⁰C temporarily. The frozen root sample

was ground into a uniform semi-liquid paste, which was then transferred into a 2mL labelled

collection tube. The other half of the sample then was ground in the same tissue grinder to a

similar consistency, then added to the same collection tube. This collection tube was then

returned to the -80⁰C freezer.

Root Cleanup

Each ground root sample was divided between 2 collection tubes (5mL) in approximately

equal halves, and 3mL nuclease-free, autoclaved, 0.2µL filtered water was added to each. A

flame-sterilized scoop was then used to add 6 glass beads to each tube. Tubes were then

disrupted so that roots would not compact at one end of the tube. Tubes were placed on a shaker

running at its maximum speed and at an angle (again, to prevent root compaction), secured with

rubber bands, and allowed to shake for 2 hours. Following this, tubes were centrifuged at 300 x

13

g for 1 minute to separate the liquid and solid contents of each tube. The liquid contents of each

tube were pipetted out and run through a 2µL filter into 2 labelled, tared collection tubes. Pipette

transfer was performed between tubes as needed to ensure that both tubes had 0.1g of liquid

contents. The resultant collection tubes were centrifuged at 15,000 x g for 15 minutes, after

which supernatant was immediately poured off and the pellet stored at -80⁰C for DNA

extraction.

DNA Extraction and Sequencing

Extraction was carried out following a slightly modified version of the Mo Bio

Powersoil® DNA Isolation Kit instructions, as explained below. The provided solution was

pipetted out of the clean, labelled PowerBead tubes and used to suspend the sample pellet before

the mixture was replaced in its original PowerBead tube along with 60µL of provided Solution

C1. PowerBead tubes were then secured and shaken on a flat-bed vortexer for 15 minutes.

Tubes were centrifuged at 10,000 x g for 30 seconds and resultant supernatant was transferred to

clean, labelled 2mL collection tubes. 250µL of provided Solution C2 was pipetted into these

tubes. The tubes were vortexed for 5 seconds and then incubated at 4⁰C for 5 minutes, after

which they were centrifuged at 10,000 x g for 1 minute. Avoiding the resultant pellet, 600µL of

supernatant was pipetted into clean, labelled 2mL collection tubes, to which 200µL of Solution

C3 was added, vortexed for 5 seconds, and incubated at 4⁰C for 5 minutes. The tubes were

centrifuged again at 10,000 x g for 1 minute. Another pellet resulted and was avoided to pipet

600µL of supernatant into a clean, labelled 5mL collection tube. 1,200µL of Solution C4 was

added and vortexed for 5 seconds.

14

At this point, the two duplicates of each sample were recombined using a Powervac ™

Manifold, passing the tubes’ contents through a spin filter column in 600µL increments. 800µL

of 100% ethanol was then passed through the spin filter, followed by 500µL of Solution C5, with

each addition being allowed to drain completely before the stopcock was turned. The vacuum

was run for an additional 1 minute to allow the spin filter membrane to dry, then turned off to

remove the spin filter columns and replace them in their original 2mL collection tubes, which

were then centrifuged at 13,000 x g for 1 minute to fully dry the membrane. The spin filter

column was then transferred to a clean, labelled 2mL collection tube, where 100µL of Solution

C6 was added to the center of the white filter membrane. The tubes were centrifuged at 10,000 x

g for 30 seconds before the spin filter was discarded and the remaining liquid was transferred to

a -80⁰C freezer until it could be primed and transferred for PCR amplification and subsequent

16S rRNA sequencing. The ultimate product was 33 samples, each combination of treatments

replicated 4 times for a total of 32 samples, plus a control where all steps were carried out as

written but without any actual root tissue (so as to correct for potential procedural contamination

and human error).

Illumina MiSeq 16S rRNA Mothur Sequence Analysis

The initial data from the sequencing process provided four replications per sample. Six

samples returned insufficient sequence data and were discarded. The discarded samples were the

procedural contamination control, as well as two reduced tillage/no-glyphosate/GR samples, one

conventional tillage/glyphosate/GR sample, one conventional tillage/glyphosate/non-GR sample,

and one conventional tillage/no-glyphosate/non-GR sample.

15

All four replicate sequencing runs for each sample were treated as a single sample during

mothur processing. Mothur is a free, Windows-compatible software intended to simplify 16s

rRNA sequence analysis [13]. The replicates were organized in a .files file which assigned the

appropriate MiSeq .fastq reads to each sample. Table 1 shows the order of mothur instructions

used to provide diversity analyses, sequence screening, and sequence alignment, among other

features [13] used to generate the experiment’s data.

Table 1. Commands used in the bioinformatics software mothur to process the 16S rRNA gene sequence data obtained in this study

Command Purposemake.contigs Reads were merged into a .fasta file of all the sequences as well as

a .groups file keeping track of which sequences came from which sample.

summary.seqs Provided a table showing basic information about the number of sequences in the current .fasta file, the number of bases in those sequences (as a percentage), the number of ambiguous bases, and indicators of poor sequence quality. This command was used repeatedly in the procedure to be certain the commands were executing correctly.

screen.seqs(maxambig=1, maxlength=275) To eliminate errors from sequences far exceeding the normal 250 base length, sequence size was limited to 275 base pairs, with 1 ambiguous base permitted.

unique.seqs Generated a .names file to merge duplicate sequences and lower processing demands.

count.seqs Generated a count table from the .names and .groups files.align.seqs(reference=silva.v4.fasta) Aligned the sequences using the SILVA V4 database. The

SILVA database included Archaea, Eukaryota, and Bacteria domains and large and small subunit rRNA genes. It further included taxonomic classifications, multiple sequence alignment, type strain information, the latest valid nomenclature, and quality checking for all sequences [10].

screen.seqs(start=1968, end=11550, maxhomop=8)

Eliminated erroneous sequences with more than 8 identical bases in a row and mis-amplifications that did not correspond to the V4 region under examination.

filter.seqs(vertical=T, trump=.) Non-informative gap sequences were filtered out, typically reducing the number of sequences under scrutiny.

unique.seqs As the editing might have caused more sequences to become identical, these were eliminated.

pre.cluster(diffs=2) Sequences so slightly different (2 bases out of 250) that they were liable to be PCR or sequencing error instead of genetic variation were pooled back together.

chimera.uchime(dereplicate=t) Listed the chimera sequences brought about by PCR error into the .accnos file via the UCHIME procedure. The dereplicate=t subcommand caused mothur to examine each sample individually.

16

remove.seqs Eliminated the listed chimera sequences.

classify.seqs(taxonomy=greengenes.tax, cutoff=80)

Using Greengenes (stored in the greengenes.fasta file), the sequences in the given files were classified and placed into Excel-compatible .taxonomy and .tax.summary files.

remove.lineage(taxon=Chloroplast-mitochondria-unknown-Archaea-Eukaryota)

Unwanted sequences such as chloroplast or mitochondrial genetic information were removed.

cluster.split(splitmethod=classify, taxlevel=4,cutoff=0.15, processors=4)

To reduce the processing time, sequences were grouped via taxonomy, and from there further clustered into operational taxonomic units (OTUs).

make.shared(label=0.03) Created an Excel-compatible .shared file which contained how frequently a sequence belonging to a particular OTU was found in a sample.

count.groups Checked how many sequences were in each sample so as to eliminate samples that had very low sequence counts from further analysis.

classify.otu Identified the established OTUs and generated Excel-compatible .taxonomy and .tax.summary files.

summary.single(subsample=t, iters=1000, calc=sobs-nseqs-coverage-invsimpsom-chao-shannon-ace)

Measured alpha diversity by subsampling from each group 1,000 times. Provided the number of OTUs, number of subsampled sequences, sampling thoroughness calculations, inverse Simpson index, Schao index, Shannon index, and SACE index.

dist.shared(subsample=t) Created a similarity matrix for samples, based on presence and abundance of OTUs.

nmds Used non-metric multidimensional scaling to visualize how similar samples were based on the presence or absence of OTUs (jclass) and in terms of abundance of OTUs (thetayc).

corr.axes(method=spearman) Using Spearman’s rank correlation coefficient, axis scores from the NMDS ordinations were correlated with the abundance of different OTUs in an Excel-compatible file.

amova* Examined whether the differences between treatments was greater than the differences within treatments.

indicator* Tested for the presence of indicator OTUs to show the different distribution of OTUs between sample groups.

*These commands required “design” files which were set up to compare the different sample groups. Three separate design.files distinguished between samples with conventional and reduced tillage, GR and non-GR plants, and glyphosate and no-glyphosate treatments in order to allow mothur to compare sample compositions. In a fourth file, total variable combinations were compared to one another.

RESULTS AND DISCUSSION

Alpha diversity analyses looked into the richness of community composition in the

individual samples [15] and these results were used to generate the beta diversity analyses. Data

17

was primarily analyzed via three different beta diversity analyses, each of which compared the

samples’ distinct microbial compositions for significant differences.

Spearman’s Correlation Coefficient and NMDS Analysis

Non-metric multidimensional scaling (NMDS) was used to represent how similar

samples were to each other. This generated a graph of which bacterial taxa appeared responsible

for the greatest differentiation between samples. In Figure 1, the most prominent OTUs are

highlighted in black while the other bacteria are denoted in yellow.

Figure 1. NMDS plot of the bacteria that appeared most responsible for the greatest differentiation between samples.

-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

The axes from Figure 1 were correlated with OTU abundance via Spearman’s rank

correlation coefficient. The distance from the NMDS graph’s origin point is the value presented

as length in Table 2. All length values calculated to be above 0.75 were considered significant

and the fifteen OTUs indicated to have the most significance were identified in Table 2

according to this length value.

18

Table 2. The fifteen most significant OTUs, as calculated by NMDS, OTU abundance, and the Spearman’s rank

correlation coefficient.

OTU Identity Length

OTU 0001 Escherichia coli 0.86907

OTU 0004 Acinetobacter guillouiae 0.648784

OTU 0005 unclassified Burkholderia 0.762337

OTU 0006 unclassified Acidovorax 0.649012

OTU 0009 unclassified Cloacibacterium 0.601773

OTU 0010 Rhizobium leguminosarum 0.710105

OTU 0037 unclassified Pseudomonadaceaef 0.743236

OTU 0076 Bosea genosp. 0.609317

OTU 0092 unclassified Sphingomonas 0.602735

OTU 0099 unclassified Aeromonadaceaef 0.677028

OTU 0136 unclassified Prevotella 0.671877

OTU 0164 Arthrobacter psychrolactophilus 0.719996

OTU 0185 unclassified Actinomycetaleso 0.677416

OTU 0262 Dongia mobilis 0.643942

OTU 0620 unclassified Koribacteraceaef 0.69042f indicates classification continued only to the family levelo indicates classification continued only to the order level

The results indicated OTUs 0001 and 0005 were the only two sequences with length

sufficient to be considered significant, suggesting only limited differences between sample

compositions. Meanwhile, OTUs 0006 and 0185 were both listed as being significant indicators

of difference in later beta diversity analyses, which potentially supports their being a relatively

large presence despite not meeting the significance requirements for this particular analysis.

OTU 0001 was also the largest presence among the samples, its OTU designation of 0001

indicating that this sequence was seen in aggregate more times than any other OTU (OTU

frequency increases inversely to OTU numeric designation).

19

AMOVA Analysis

Exploration of the differences between samples continued with analysis of molecular

variance (AMOVA; Table 3).

Table 3. Differences between treatments as determined from AMOVA analysis. Variables are abbreviated as

glyphosate treatments (GLY), no glyphosate (nGLY), reduced tillage (RT), conventional tillage (CT), GR (same),

and non-GR (nGR).

Sample Groups Compared p-value

All possible combinations <0.001*

CT and RT 0.015*

nGR and GR 0.031*

CT/nGLY/nGR and CT/nGLY/GR 0.037

CT/nGLY/GR and RT/nGLY/nGR 0.063

CT/nGLY/GR and CT/GLY/nGR 0.076

CT/nGLY/nGR and CT/GLY/GR 0.087

CT/GLY/GR and RT/noGLY/nGR 0.088

CT/nGLY/nGR and RT/nGLY/nGR 0.094

CT/GLY/nGR and CT/GLY/GR 0.096

CT/GLY/nGR and RT/GLY/nGR 0.099

CT/nGLY/nGR and CT/GLY/nGR 0.1

CT/GLY/nGR and RT/nGLY/nGR 0.102

CT/GLY/GR and RT/nGLY/GR 0.104

CT/nGLY/nGR and RT/GLY/GR 0.105

CT/nGLY/nGR and RT/GLY/nGR 0.107

CT/GLY/GR and RT/GLY/GR 0.115

CT/nGLY/GR and RT/nGLY/GR 0.128

CT/GLY/nGR and RT/nGLY/GR 0.199

CT/GLY/GR and RT/GLY/nGR 0.208

CT/nGLY/nGR and RT/nGLY/GR 0.208

RT/nGLY/nGR and RT/nGLY/GR 0.216

RT/nGLY/nGR and RT/GLY/GR 0.295

CT/nGLY/GR and CT/GLY/GR 0.341

20

RT/nGLY/GR and RT/GLY/GR 0.364

CT/GLY/nGR and RT/GLY/GR 0.388

CT/nGLY/GR and RT/GLY/GR 0.449

RT/nGLY/nGR and RT/GLY/nGR 0.45

RT/nGLY/GR and RT/GLY/nGR 0.571

CT/nGLY/GR and RT/GLY/nGR 0.58

GLY and nGLY 0.69

RT/GLY/nGR and RT/GLY/GR 0.844

*indicates p-value was considered significant in AMOVA

A simultaneous comparison of all eight combinations of variables led to p<0.001, which

was significant; therefore the differences between all eight sample groups exceeded the variation

within the samples themselves. This indicated that, as a group, the variables studied here

(tillage, glyphosate application, and transgenic corn type) did have an effect on the bacterial

endophyte community composition. However, when sample treatments were compared

individually to one another, AMOVA was unable to significantly detect differences between any

individual pairs of treatments. When variables were strictly compared only to their counterpart,

AMOVA was also unable to detect significant differences between GLY and nGLY. It did

detect significant differences between GR and nGR corn plants, (p = 0.031), and between RT and

CT (p = 0.015). This indicated that tillage approach and genetically modified plants (compared

to non-modified) both led to a significant difference in the endophyte community.

Indicator OTU Analysis

Sequences were tested for indicator OTUs (those which differed significantly between

samples; Table 4).

21

Table 4. Indicator OTUs displayed with their identities as far as the greengenes database could classify their

sequences, the samples they were predominant within, and p-values for the significance of these differences.

OTU Identification Presence P-value

0006 unclassified Acidovorax CT/nGLY/nGR 0.047

0029 unclassified Kaistobacter RT 0.003

0029 unclassified Kaistobacter RT/nGLY/GR 0.044

0035 unclassified Sinobacteraceaef RT/nGLY/GR 0.043

0036 Lechevalieria aerocolonigenes GR 0.039

0048 Prevotella melaninogenica CT/nGLY/nGR 0.048

0055 Rothia mucilaginosa CT/nGLY/nGR 0.044

0063 unclassified Kaistobacter GR 0.035

0064 Sphingomonas wittichii RT 0.025

0065 unclassified Chryseobacterium nGLY 0.044

0065 unclassified Chryseobacterium CT/nGLY/nGR 0.024

0080 unclassified Prevotella CT/nGLY/nGR 0.045

0088 unclassified Alicyclobacillus GR 0.037

0096 unclassified Gaiellaceaef RT/nGLY/GR 0.024

0102 Streptomyces reticuliscabiei GR 0.042

0139 unclassified Actinomycetaleso RT/nGLY/GR 0.013

0144 unclassified Solibacillus CT/nGLY/nGR 0.038

0169 unclassified Lachnospiraceaef CT/nGLY/nGR 0.048

0180 unclassified Clostridium CT/nGLY/nGR 0.046

0183 unclassified Candidatus Xiphinematobacter RT/nGLY/GR 0.007

0185 unclassified Actinomycetaleso CT/nGLY/nGR 0.019

0190 unclassified Prevotella nGR 0.036

0209 unclassified Intrasporangiaceaef RT 0.047

0216 unclassified Streptococcus CT/nGLY/nGR 0.049

0289 unclassified Actinomycetaleso RT/nGLY/GR 0.012

0385 unclassified Neisseriaceaef CT/nGLY/nGR 0.048

0442 unclassified Chitinophagaceaef RT/nGLY/GR 0.017

0475 unclassified iii1-15o RT/nGLY/GR 0.01

0480 unclassified Catellatospora RT/nGLY/GR 0.011

0726 unclassified Chitinophagaceaef RT/nGLY/GR 0.012

0739 unclassified Ruminococcus CT/nGLY/nGR 0.047

0774 unclassified Gammaproteobacteria RT/nGLY/GR 0.01

0875 unclassified Coprococcus CT/GLY/nGR 0.043

22

0882 unclassified Gaiellaceaef RT/nGLY/GR 0.01f means classified only to family levelo means classified only to the order level

The only significant difference between the GLY and nGLY treatments was OTU 0065

(unclassified Chryseobacterium), which was found to be more prevalent in nGLY samples. This

same OTU was also present in CT/nGLY/nGR samples (a specific subtype of nGLY sample).

As none of the other OTUs from that variable combination were noted during this particular

comparison, it can be assumed that either OTU 0065 was particularly prevalent in other nGLY

samples compared to the other OTUs in CT/nGLY/nGR, or the other significant OTUs from

CT/nGLY/nGR were sufficiently present in GLY samples to conceal their prevalence in nGLY

samples. It is important to note that the RT/nGLY/GR sample mentioned in the methods may

have affected the GLY/nGLY comparison results, though this effect would be mitigated by many

other GLY and nGLY samples.

When the samples were divided by tillage treatment, OTUs 0029 (unclassified

Kaistobacter), 0064 (Sphingomonas wittichii), and 0209 (unclassified Intrasporangiaceae) were

significantly present in RT samples; no OTUs were significantly more prevalent in the CT

samples. OTU 0029 (and none of the others) was also represented as significant in a specific

variable combination group (RT/nGLY/GR), indicating particularly high levels within that

group. This group was mentioned before as returning two samples with no data, and these

results are accordingly inconclusive. Additionally, OTU 0209’s significance may be the

overemphasis of a rare bacterium due to a comparatively rare presence in the data set. Overall,

however, the presence of multiple significantly different OTUs between CT and RT samples

supports the AMOVA suggestion of significant difference due to tillage treatments; these OTUs

can be assumed to be what that difference was derived from.

23

In terms of GR and nGR corn, OTUs 0036 (Lechevalieria aerocolonigenes), 0063

(unclassified Kaistobacter), 0088 (unclassified Alicyclobacillus), 0102 (Streptomyces

reticuliscabiei), and 0190 (unclassified Prevotella) all differed significantly. All but OTU 0190

were present predominantly in GR corn. Overall, the presence of these OTUs again reflects the

AMOVA findings. The risk of slightly biased data from the RT/nGLY/GR sample is likely

compensated for by the many other GR and nGR samples. Of some concern is the significance

of OTU 190, an unidentified Prevotella, a genus of gram-negative, obligately anaerobic,

nonmotile, nonsporeforming, pleomorphic rods, that contains several species found in the oral

cavity [16]; accordingly, it may have resulted from contamination. Both it and OTU 102’s

relatively smaller presence in the samples may suggest that they were the product of

overemphasis.

When samples were compared by variable combinations, two combinations were

revealed to be especially unique. The first was CT/nGLY/nGR samples, which showed OTUs

0006 (unclassified Acidovorax), 0048 (Prevotella melaninogenica), 0055 (Rothia mucilaginosa),

0065 (unclassified Chryseobacterium), 0080 (unclassified Prevotella), 0144 (unclassified

Solibacillus), 0169 (unclassified Lachnospiraceae), 0180 (unclassified Clostridium), 0185

(unclassified Actinomycetales), 0216 (unclassified Streptococcus), 0385 (unclassified

Neisseriaceae), and 0739 (unclassified Ruminococcus) to be present significantly in comparison

to the rest of the endophyte samples. OTU 0006 and 0185 both were distinguished in the NMDS

analysis as having potentially significant presences in the data set. No other OTUs from the

NMDS analysis were seen in the indicator OTUs analysis results, lending these two a special

significance and accordingly making the CT/nGLY/nGR variable combination appear to have a

drastic effect on endophyte microbial community composition. OTU 0169’s Lachnospiraceae

24

family typically contains gastrointestinal bacteria in humans and ruminants, all of which are

strict anaerobes and primarily nonsporeforming. Its members can also occur in the human oral

cavity [8] and its presence may indicate contamination. Likewise, OTU 0055, Rothia

mucilaginosa, is an upper respiratory bacteria that causes opportunistic infections in severely

immunocompromised human hosts [7], and almost certainly resulted from contamination, but the

other prokaryotes have more ambiguous origins. OTU 0065 was mentioned in previous

comparisons (nGLY and GLY), which may suggest that the CT/nGLY/nGR combination was

affected by that comparison or vice versa. Any of the OTUs 0144, 0169, 0180, 0185, 0216,

0385, and 0739 could be rare, overemphasized bacteria. Conversely, it is possible that the

combinations of one variable in conjunction with another may have had some holistic effect

upon the endophyte composition, affecting OTU 0065 as well as others (especially OTU 0006

and 0185 from the NMDS analysis).

The second significant variable combination referred to RT/nGLY/GR samples, which,

due to low returns, only had two samples of the four intended for this experiment. Accordingly

any results are unreliable and are less likely to be indicative of actual differences in the

endophyte community composition than other comparisons. OTUs 0029 (unclassified

Kaistobacter), 0035 (unclassified Sinobacteraceae), 0096 (unclassified Gaiellaceae), 0139

(unclassified Actinomycetales), 0183 (unclassified Candidatus Xiphinematobacter), 0289

(unclassified Actinomycetales), 0442 (unclassified Chitinophagaceae), 0475 (unclassified iii1-

15), 0480 (unclassified Catellatospora), 0726 (unclassified Chitinophagaceae), 0774

(unclassified Gammaproteobacteria), and 0882 (unclassified Gaiellaceae) were significantly in

this group. As with the CT/nGLY/nGR sample above, none of these OTUs except for 0029 were

seen in the previous analyses, indicating either overemphasis of rare microbes or a potential

25

holistic effect of these combined variables. In the case of this variable combination, more study

would be required to confirm or deny these suppositions, but as only three of its OTUs have

designations less than one hundred, it is likely that several of the OTUs described were not so

significant as they appeared.

A third variable combination (CT/GLY/nGR) had OTU 0875 (unclassified Coprococcus)

represented to a significantly heightened degree. This may reflect another holistic effect of the

variable combinations, albeit a much more subtle one, or the previously discussed overemphasis

of rare bacteria, which is more likely as the OTU designation of 0875 indicates the bacteria to be

present in relatively low numbers.

CONCLUSION

Although the prevalence of E.coli indicates that there was likely contamination, its lack

of appearance as a significant OTU in any of the sample groups suggests that the contamination

in question was reasonably uniform, and thus cannot invalidate the other results. The

combination of NMDS, AMOVA, and OTU indicator testing indicate that tillage treatments and

GMO corn plants both likely influence prokaryotic endophyte community composition. Whether

or not glyphosate treatments have an effect is less certain, and if so, it appears to be a rather

subtle one (at least within the time frame of this experiment). There also was a suggestion of

variables combining to have net effects on the endophyte community composition, particularly

when dealing with CT/nGLY/GR (and also possibly with RT/nGLY/GR, though the unreliability

of the latter combination makes it impossible to say within the scope of this experiment). These

results suggest that to understand the full nature of endophyte communities, it will not be enough

26

to simply select straightforward variables; combinations of variables may affect each other to a

significant degree. There is also potential to discover unanticipated effects of these variables on

the endophyte community composition over periods of time longer than one year, especially

given climatological variations and the potential consequences therein.

Many of the OTUs considered significant were unable to be identified by the current

databases. As so many of these prokaryotes remain unknown, it is difficult to speculate as to the

broader effects of any of the variables examined in this experiment on the agricultural process or

the ecological ramifications. It may prove possible to derive some understanding from

examining phylogenetic data, especially given the improvements in accurate prokaryote

classification over the recent years and optimally, the field of microbiology will only continue to

grow. The continuing expansion of the databases will allow future experiments with endophytes

to yield more telling results.

Larger sample sizes could also be used in the future, to minimize the risk of procedural

error and the consequences of data loss. Though it is currently beyond the scope of this

experiment, future studies could compare the endophyte communities of plants other than Zea

mays, and hopefully such studies will pave the way to a broader understanding of prokaryotes,

their role in our ecosystem as a whole, and the application of the latest agricultural techniques.

Finally, to address the issue of contamination, future endophyte research should continue to

modify the experimental protocol at any given level in order to better isolate the desired bacterial

endophyte rRNA and reduce contamination.

27

BIBLIOGRAPHY

1. Boddey, R. M., Urquiaga, S., Reis, V., and Dobereiner, J. 1991. Biological nitrogen

fixation associated with sugar cane. Plant & Soil. 137: 111-117.

2. Caporaso, G. J., Lauber, C. L., Walters, W. A., Berg-Lyons, D., Huntley, J., Fierer,

N., Owens, S. M., Betley, J., Fraser, L., Bauer, M. et al. 2012. Ultra-high-throughput

microbial community analysis on the Illumina HiSeq and MiSeq platforms. The ISME

Journal. 6: 1621-1624.

3. Carpenter-Boggs, L., Stahl, P. D., Lindstrom, M. J., and Schumacher, T. E. 2003.

Soil microbial properties under permanent grass, conventional tillage, and no-till

management in South Dakota. Soil & Tillage Research. 71: 15-23.

4. Duke, S. O. and Powles, S. B. 2008. Glyphosate: a once-in-a-century herbicide. Pest

Management Science. 64: 319-325.

5. Groenigen, K., Bloem, J., Baath, E., Boeckx, P., Rousk, J., Bodé, S., Forristal, D.,

and Jones, M. B. 2010. Abundance, production and stabilization of microbial biomass

under conventional and reduced tillage. Soil Biology & Biochemistry. 42: 48-55.

6. Hirsch, G. and Braun, U. 1992. Communities of parasitic microfungi. In: W.

Winterhoff editors. Handbook of Vegetation Science. Vol 19, Fungi in Vegetation

28

Science. Kluwer Academic Dordrecht, p. 225-250.

7. Lee, A. B., Harker-Murray, P., Ferrieri, P., Schleiss, M. R., and Tolar, J. 2007.

Bacterial meningitis from rothia mucilaginosa in patients with malignancy or undergoing

hematopoietic stem cell transplantation. Pediatric Blood & Cancer. 50(3): 673-676.

8. Meehan, C. J. and Beiko, R. G. A phylogenetic view of ecological specialization in the

Lachnospiraceae, a family of digestive tract-associated bacteria. 2014. Genome Biology

and Evolution. 6(3): 703-713.

9. Padgette, S.R., Kolacz, K. H., Delannay, X., Re, D. B., LaVallee, B. J., Tinius, C. N.,

Rhodes, W. K., Otero, Y. I., Barry, G. F., Eichholtz, D. A. et al. 1995. Development,

Identification, and Characterization of a Glyphosate-Tolerant Soybean Line. Crop

Science. 35(5): 1451-1459

10. Quast, C., Pruesse, E., Yilmaz, P., Gerken, J., Schweer, T., Yarza, P., Peplies, J., and

Glöckner, F. O. 2012. The SILVA ribosomal RNA gene database project: improved

data processing and web-based tools. Nucleic Acids Research. 41: D590-D596.

11. Richard, S., Moslemi, S., Sipahutar, H., Benachour, N., and Seralini, G. 2005.

Differential Effects of Glyphosate and Roundup on Human Placental Cells and

Aromatase. Environmental Health Perspectives. 113(6): 716-720.

12. Schloss, P. D. 2010. The effects of alignment quality, distance calculation method,

sequence filtering, and region on the analysis of 16S rRNA gene-based studies. PLOS

Computational Biology. 6(7): e1000844. doi: 10.1371/journal.pcbi.1000844.

13. Schloss, P. D., Westcott, S. L., Ryabin, T., Hall, J. R., Hartmann, M., Hollister, E.

B., Lesniewski, R. A., Oakley, B. B., Parks, D. H., Robinson, C. J. et al. 2009.

Introducing mothur: open-source, platform-independent, community-supported software

29

for describing and comparing microbial communities. Applied and Environmental

Microbiology. 75(23): 7537-7541.

14. Schmidt, T. M., DeLong, E. F., and Pace N. R. 1991. Analysis of a marine

picoplankton community by 16S rRNA gene cloning and sequencing. Journal of

Bacteriology. 173(14): 4371-4378.

15. Sepkoski, J. J. 1988. Alpha, beta, or gamma: where does all the diversity go?

Paleobiology. 14(3): 221-234.

16. Shah, H. N. and Collins, D. M. 1990. Prevotella, a new genus to include bacteroides

melaninogenicus and related species formerly classified in the genus bacteroides.

International Journal of Systemic Bacteriology. 40(2): 205-208.

17. Stone, J. K., Bacon, C. W., and White, J. F., Jr. 2000. An overview of endophytic

microbes: endophytism defined. In Bacon, Charles W. and White, James F., Jr. editors.

Microbial Endophytes. Marcel Dekker, New York, p. 3-20.

18. Woo, P.C.Y., Lau, S.K.P., Teng, J.L.L., Tse, H., and Yuen, K.Y. 2008. Then and

now: use of 16S rDNA gene sequencing for bacterial identification and discovery of

novel bacteria in clinical microbiology laboratories. Clinical Microbiology and Infection.

14(10): 908-934.

30